Martensit

Martensitic Transformations in Steels
A 3D Phase-eld Study
HEMANTHA KUMAR YEDDU
Doctoral Thesis
Stockholm, Sweden 2012
ISRN KTH/MSE12/12SE+METO/AVH
ISBN 978-91-7501-388-6
Materialvetenskap
KTH
SE-100 44 Stockholm
SWEDEN
Akademisk avhandling som med tillstnd av Kungl Tekniska hgskolan framlgges
till oentlig granskning fr avlggande av teknologie doktorsexamen i material-
vetenskap fredagen den 15 juni 2012 klockan 10.00 i sal B2, Brinellvgen 23, Ma-
terialvetenskap, Kungl Tekniska hgskolan, 10044 Stockholm.
Hemantha Kumar Yeddu, June 2012
Tryck: Universitetsservice US AB
iii
The highest education is that which does not merely give us informa-
tion but makes our life in harmony with al l existence. Rabindranath
Tagore, Nobel Laureate (1913).
v
Dedicated to the most wonderful persons in my life Mom and Aunty
vii
Abstract
Martensite is considered to be the backbone of the high strength of many
commercial steels. Martensite is formed by a rapid diusionless phase trans-
formation, which has been the subject of extensive research studies for more
than a century. Despite such extensive studies, martensitic transformation
is still considered to be intriguing due to its complex nature. Phase-eld
method, a computational technique used to simulate phase transformations,
could be an aid in understanding the transformation. Moreover, due to the
growing interest in the eld of Integrated computational materials engineer-
ing (ICME), the possibilities to couple the phase-eld method with other
computational techniques need to be explored.
In the present work a three dimensional elastoplastic phase-eld model,
based on the works of Khachaturyan et al. and Yamanaka et al., is developed
to study the athermal and the stress-assisted martensitic transformations oc-
curring in single crystal and polycrystalline steels. The material parameters
corresponding to the carbon steels and stainless steels are considered as input
data for the simulations. The input data for the simulations is acquired from
computational as well as from experimental works. Thus an attempt is made
to create a multi-length scale model by coupling the ab-initio method, phase-
eld method, CALPHAD method, as well as experimental works. The model
is used to simulate the microstructure evolution as well as to study various
physical concepts associated with the martensitic transformation.
The simulation results depict several experimentally observed aspects as-
sociated with the martensitic transformation, such as twinned microstructure
and autocatalysis. The results indicate that plastic deformation and auto-
catalysis play a signicant role in the martensitic microstructure evolution.
The results indicate that the phase-eld simulations can be used as tools to
study some of the physical concepts associated with martensitic transforma-
tion, e.g. embryo potency, driving forces, plastic deformation as well as some
aspects of crystallography. The results obtained are in agreement with the
experimental results.
The eect of stress-states on the stress-assisted martensitic microstructure
evolution is studied by performing dierent simulations under dierent load-
ing conditions. The results indicate that the microstructure is signicantly
aected by the loading conditions. The simulations are also used to study
several important aspects, such as TRIP eect and Magee eect. The model
is also used to predict some of the practically important parameters such as
M
s
temperature as well as the volume fraction of martensite formed.
The results also indicate that it is feasible to build physically based multi-
length scale model to study the martensitic transformation. Finally, it is
concluded that the phase-eld method can be used as a qualitative aid in
understanding the complex, yet intriguing, martensitic transformations.
Key words: Phase-eld method, Martensitic transformations, Plastic de-
formation, Multi-length scale modeling, Microstructure, Stress states, Steels.
Preface
The growing interest in the eld of computational materials engineering has lead
to a successful exploitation of computational tools to model several phase trans-
formations, especially the complex and intriguing martensitic transformation. The
present thesis deals with the study of martensitic transformations in steels using
three dimensional phase-eld simulations.
Some of the important physical concepts of martensite are explained in the
introduction part and the readers are urged to refer to the cited works for more
details. A summary of the mathematical framework of the 3D phase-eld model as
well as the modications made to it to study dierent aspects of the martensitic
transformation is presented. Finally, a summary of the results obtained from the
above mentioned works is also presented. The full details can be found in the
following appended supplements that will be referred to, throughout the thesis,
with the respective supplement numbers.
1. H.K. Yeddu, A. Malik, J. gren, G. Amberg, A. Borgenstam, Three-
dimensional phase-eld modeling of martensitic microstructure evolution in
steels, Acta Materialia 60 (2012) 1538-1547.
2. H.K. Yeddu, A. Borgenstam, J. gren, Stress-assisted martensitic transfor-
mations in steels A 3D phase-eld study, Submitted manuscript.
3. H.K. Yeddu, V.I. Razumovskiy, A. Borgenstam, P.A. Korzhavyi, A.V. Ruban,
J. gren, Multi-length scale modeling of martensitic transformations in stain-
less steels, Submitted manuscript.
4. A. Malik, H.K. Yeddu, G. Amberg, A. Borgenstam, J. gren, Three dimen-
sional elasto-plastic phase-eld simulation of martensitic transformation in
polycrystal, Submitted manuscript.
5. H.K. Yeddu, A. Borgenstam, J. gren, Eect of martensite embryo potency
on the martensitic transformations in steels A 3D phase-eld study, Journal
of Alloys and Compounds, In Press, doi:10.1016/j.jallcom.2012.01.087.
6. H.K. Yeddu, A. Borgenstam, P. Hedstrm, J. gren, A phase-eld study
of the physical concepts of martensitic transformations in steels, Materials
Science and Engineering A 538 (2012) 173-181.
ix
x PREFACE
Respondents contribution to the papers
Supplement-1: The respondent performed all the work and wrote the entire
paper with inputs from J. gren (J..), A. Borgenstam (A.B.), A. Malik
(A.M.) and G. Amberg (G.A.).
paper with inputs from J.. and A.B.
Supplement-3: The respondent performed all the phase-eld part of the work
as well as wrote major part of the paper with inputs from J.. and A.B. The
ab initio part of the work is performed by V.I. Razumovskiy (V.I.R.) with
inputs from P.A. Korzhavyi (P.A.K.) and A.V. Ruban (A.V.R.).
Supplement-4: The respondent contributed to the model development as well
as to drafting the paper with inputs from J.., A.B., A.M. and G.A.
paper with inputs from J.. and A.B.
paper with inputs from J.., A.B. and P. Hedstrm (P.H.).
This work has also resulted in the following presentations
Invited talk on Role of Plasticity during martensitic microstructure evolution
in steels: A 3D phase eld study, presented at Plasticity 2012, San Juan,
PR, USA, January 2012.
Invited talk on Phase Transformations in Materials - A mesoscale study
on steels using phase-eld approaches, presented at Los Alamos National
Laboratory, Los Alamos, USA, January 2012.
Oral presentation on 3D Phase Field Modeling of Martensitic Microstructure
Evolution in Steels, presented at Phase Transformations Conference (PTM
2010), Avignon, France, June 2010.
Poster presentation on Physical Parameters Determination From 3D Phase
Field Simulations of Martensitic Transformations in Steels, presented at
TMS Annual Meeting, San Diego, USA, February 2011.
Poster presentation on 3D Modeling of Martensitic Transformations in Steels
by Integrating Thermodynamics, Phase Field Modeling and Experiments,
at 1st World Congress on Integrated Computational Materials Engineering
(ICME), Seven Springs, PA, USA, July 2011.
Poster presentation on Eect of martensite nucleus potency on the marten-
sitic transformation in steels: A Phase Field study, presented at ICOMAT
2011, Osaka, Japan, September 2011.
Contents
Preface ix
Contents xi
1 Introduction 1
1.1 Aim of the present work . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Martensite 3
2.1 Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Plastic deformation aspects . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 Nucleation and growth . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.5 Thermodynamic aspects . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.6 Crystallography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3 Phase-eld method 13
4 Phase-eld modeling of martensitic transformations 17
4.1 Morphology and transformation . . . . . . . . . . . . . . . . . . . . . 18
4.3 Nucleation and growth aspects . . . . . . . . . . . . . . . . . . . . . 24
5 Summary of results 27
5.1 Morphology and transformation . . . . . . . . . . . . . . . . . . . . . 27
5.3 Nucleation and growth aspects . . . . . . . . . . . . . . . . . . . . . 38
6 Concluding remarks and future prospects 47
xi
xii CONTENTS
Acknowledgements 51
Bibliography 53
Chapter 1
Introduction
Steels, due to their superior mechanical properties such as good formability and
high strength, have become a vital class of materials that are employed in a wide
range of applications, ranging from kitchen cutlery to spacecraft. Based on the
alloying composition, steels can be broadly classied as: carbon steels, low alloy
steels and stainless steels. Carbon steels, whose basic alloying elements are iron
and carbon, are the most commonly used steels for general applications and hence
contribute to almost 90% of the world steel production. Stainless steels, whose basic
alloying elements are iron, chromium and nickel, are employed for special type of
applications that demand good corrosion resistance along with the above mentioned
mechanical properties. Based on the main constituent phases, i.e. austenite, ferrite,
martensite and mixture of austenite and ferrite, steels can be broadly classied as:
Austenitic, Ferritic, Martensitic and Duplex steels, respectively. The mechanical
properties of steels are determined by their microstructure and the constituent
phases and hence the study of structure-property relations of steels is essential for
both the industry as well as the academia.
The discovery that the hardest steels contained an acicular microstructural con-
stituent had been a pioneering eort, probably by Sorby [1], in explaining the cause
for the widely diering mechanical properties of steels. The acicular structure was
named as martensite, in honour of the German metallographer Adolf Martens.
Martensite has since then been inevitably associated with the high strength of
many commercial steels. Martensite () forms from austenite () in a diusionless
manner, i.e. there is no variation in composition of the two solid phases during
the transformation, and hence the phase transformation can be classied as a solid
state diusionless phase transformation, named martensitic transformation. Due to
its practical importance, the martensitic transformation has undergone extensive
studies that have lead to coining the term martensitic transformation as a generic
name to describe a solid state diusionless phase transformation, with specic fea-
tures, that occurs in a wide variety of materials, such as steels, shape memory alloys
as well as in ceramics.
1
2 CHAPTER 1. INTRODUCTION
Based on the mode of formation, martensitic transformation in steels can be
classied as: athermal, i.e. rapid transformation during quenching, isothermal, i.e.
slower transformation while holding the steel at a constant temperature, stress-
assisted, i.e. by application of stress below the yield limit of the steel, and strain-
induced, i.e. by plastic deformation of the steel. Thus martensite can form in the
steel products during their application. Some special steels, e.g. TRIP steels, are
known for their ability to enhance their mechanical properties during in-service
by means of martensite formation under applied stress. Hence in order to better un-
derstand the microstructure-property relations of steels, it is essential to thoroughly
understand the martensite formation under various thermo-mechanical conditions.
Though the martensite formation has been studied quite extensively over the
past century, it is still very intriguing due to its complex nature that leaves some
questions unanswered. The progress made over the past two decades in the eld
of electron microscopy as well as in the eld of in situ studies using synchrotron
radiation seems promising for performing in situ studies of the martensitic trans-
formation. However the rapidity of the transformation makes it very dicult to be
studied by in situ experiments.
Phase-eld method, a computational technique, has proved to be successful to
simulate the microstructure evolution during various phase transformations [24],
including martensitic transformation [514]. Hence an in-depth study of martensitic
transformation using phase-eld method, especially in 3D, could aid in understand-
ing the transformation. Moreover, due to the growing interest in the Integrated
computational materials engineering (ICME), i.e. materials design and optimiza-
tion by means of a combination of multi-length scale modeling approaches, the
possibilites to couple the phase-eld technique with other computational techniques
need to be explored.
1.1 Aim of the present work
In the present work, a three dimensional elastoplastic phase-eld model is devel-
oped to study the athermal martensitic transformation, occurring in single crystal
and polycrystalline steels, as well as the stress-assisted martensitic transformation
occurring in single crystal steels. 3D phase-eld simulations are performed by con-
sidering the material parameters corresponding to the carbon steels and stainless
steels as the input data, which is acquired from computational as well as from
experimental works. Thus an attempt is made to create a physically based multi-
length scale model by coupling ab-initio method at the atomistic level, phase-eld
method at the mesoscopic level, experimental works at the macro level and CAL-
PHAD method, which is based on the fundamental principles of thermodynamics.
The model is used to study the microstructure evolution as well as to study vari-
ous physical aspects, such as morphology and transformation, plastic deformation,
nucleation and growth, thermodynamics as well as some aspects of the crystallog-
raphy, associated with the martensitic transformation.
Chapter 2
Martensite
2.1 Morphology
Depending on the alloy composition, martensite forms in the shape of laths, i.e.
ruler shaped units, or plates, i.e. lenticular shaped units, as shown in Fig. 2.1.
Carbon content of the alloy plays a major role in determining the martensite mor-
phology. It has been experimentally observed that martensite forms in the shape
of laths in low carbon steels and in the shape of plates in high carbon steels [15].
Figure 2.1: Light optical microscope images of the microstructure of (a) lath
martensite in an IF steel with very low carbon content of 0.0049 wt% C [16] (b)
plate martensite in Fe1.86 wt% C alloy [17].
In the case of lath martensite, several parallel martensite units are formed ad-
jacent to each other [15]. It has been reported that lath martensite forms in blocks
and packets in an austenite grain as shown in Fig. 2.1 [16, 18]. The strength and
toughness of martensitic steels, with lath morphology, are strongly related to packet
3
4 CHAPTER 2. MARTENSITE
and block sizes [18, 19].
In the case of plate martensite, non-parallel martensite units are formed [15].
Moreover, a distinct feature in the case of plate martensite is the midrib, which is
reported to be the rst forming unit consisting of many transformation twins [20,21].
On either side of the midrib there exist twinned regions, with some twins that extend
from the midrib, and untwinned regions, with dislocations [20, 21].
2.2 Transformation
During rapid cooling (quenching) of austenite, athermal martensitic transformation
begins at the martensite start temperature (M
s
). Thereafter, the volume fraction
of martensite increases with decreasing temperature and nally, the transformation
is completed on reaching the martensite nish temperature (M
f
). In the case of
stress assisted martensite, the martensite start temperature at a given stress level
is termed M
s
.
Fig. 2.2 shows the microstructural images obtained during an in situ observa-
tion, by using laser scanning confocal microscope, of plate martensite formation by
means of athermal transformation during rapid quenching of a high carbon steel
with a composition of Fe-0.88%C-4.12 %Cr, expressed in mass % [22]. The real
time and temperature can be seen in the upper left corner. Slightly above the
M
s
, one can see untransformed austenite grains in Fig. 2.2a. Slightly below the
M
s
, martensitic transformation initiates in a heterogeneous manner, i.e. nucleation
occurs at dierent sites as can be seen in Fig. 2.2b. One can see, in Fig. 2.2b,
the untransformed austenite grains as well as partly transformed austenite grains
with martensite plates. At a temperature close to M
f
, one can see an almost
completely transformed steel with complex martensitic microstructure containing
several martensite plates oriented in dierent directions, as shown in Fig. 2.2c.
Thus it can be understood that the martensite volume fraction increases with de-
creasing temperature.
During rapid cooling, a diusion controlled transformation of austenite to fer-
rite, with a very low carbon content, does not occur due to lack of time. However
as the temperature is reduced below the T
0
temperature, i.e. the temperature
where the Gibbs energies of ferrite and austenite are the same, there exists a ther-
modynamic driving force available for the formation of ferrite (martensite) with
the same composition as that of austenite, i.e. occurrence of a diusionless phase
transformation.
Martensitic transformation leads to the crystallographic transformation of face
centered cubic (FCC) austenite in to body centered cubic (BCC) martensite. The
carbon atoms that are randomly distributed on the interstitial sites in FCC do
not have time to migrate to the BCC in a random manner and hence move in a
coordinated motion. This increases the tetragonality of the BCC lattice and thus
the carbon containing martensite is of body centered tetragonal (BCT) structure.
The tetragonality of martensite increases with increasing carbon content [23].
2.2. TRANSFORMATION 5
Figure 2.2: In situ observation, by using laser scanning confocal microscope, of
plate martensite formation in a high carbon steel, where the real transformation
time and temperature are marked in the upper left corner [22]. (a) Microstructure
at a temperature slightly above M
s
. (b) Microstructure at slightly below M
s
. (c)
Final martensitic microstructure at a temperature close to M
f
.
In addition to the increased tetragonality, the increase in carbon content also
leads to a volume expansion, i.e. dilatation. Moreover, the solid state nature of the
transformation, that takes place by a cooperative movement of atoms, requires the
two phases to be highly coherent and hence gives rise to a large amount of internal
stresses inside the material. The internal stresses are relaxed by means of plastic
deformation, i.e. by shear deformation. Moreover, the shear deformation needs
to be a lattice-invariant shear such that it does not give rise to a macroscopically
inhomogeneous crystal structure. Hence the lattice-invariant shear should only be
microscopically inhomogeneous [23].
Theoretically, it has also been shown that the dilatation and shear needs to be
associated with the rotation of martensite such that the combination of all three
mechanisms gives rise to the experimentally observed crystallographic orientation
relationships of martensite [23]. Thus based on the above discussion, it can be un-
derstood that the martensitic transformation occurs by a cooperative short range
movement of atoms such that the crystal lattice coherency (continuity) is main-
tained.
2.3 Plastic deformation aspects
As mentioned in Section 2.2, plastic (shear) deformation in the surrounding austen-
ite and internally in the martensite facilitates the relaxation (accommodation) of
stresses, which are generated due to the volume and shape changes caused by
the athermal transformation. In the case of stress-assisted transformation, the
plastic accommodation process is popularly known as the Greenwood-Johnson ef-
fect [2426], which can be considered to be at the origin of TRIP (Transformation
induced plasticity) phenomenon. The TRIP phenomenon has been found to be
advantageous in increasing the strength, imparted by martensite, and in increasing
the ductility, caused due to the dislocations generated by the plastic deformation
during martensitic transformation. Thus plastic deformation plays an important
role in the martensitic transformation.
The platic (shear) deformation should not disturb the overall crystal structure
and hence needs to be a lattice-invariant shear, as explained in Section 2.2. The
lattice-invariant shear deformation can be generated by the movement of perfect
dislocations or partial dislocations that gives rise to slip or twinning respectively
[23], as shown in Fig. 2.3. Fig. 2.3a corresponds to a case when there is no plastic
deformation inside the martensite and hence in this case the shape changes are
completely accommodated by deformation of the surrounding austenite. Figs. 2.3b
and 2.3c show two dierent cases of plastic deformation, i.e. slip and twinning
respectively, inside martensite.
It has been reported that the type of shearing mechanism also aects the mor-
phology of martensite [15,23]. In low carbon steels, where lath martensite is formed,
slip mechanism is preferred and in high carbon steels, where plate martensite is
formed, twinning mechanism is preferred [23]. Thus it can be understood that
lath martensite is a dislocation-rich structure [15, 23, 27], whereas plate marten-
site contains several internal twins [15, 23], as mentioned in Section 2.1. The high
dislocation density of lath martensite also indicates that the major plastic defor-
mation occurs inside martensite, despite its higher yield strength compared to that
of austenite. Dislocations also play a dominant role during nucleation and growth
of martensite, as explained in the following section.
2.4 Nucleation and growth
Martensite nucleation occurs in a heterogeneous manner [28], as can be seen in Fig.
2.2b. The nucleation initiates at a point in the austenite where there is a dense
stacking of dislocation arrays [2830]. The heterogeneous nature of nucleation can
2.5. THERMODYNAMIC ASPECTS 7
(a)
(b)
(c)
Figure 2.3: Schematic of shear deformation mechanism in martensite [23]. (a)
No lattice-invariant shear in martensite except the shape change caused by the
transformation. (b) Slip. (c) Twinning.
be attributed to the potency of the martensite embryos, which exist in a frozen-in
state and attain potency on reaching the martensite start temperature M
s
[31]. It
should also be emphasized that only the most potent embryos transform at M
s
,
whereas the less potent embryos will transform at a temperature lower than M
s
, as
and when they attain potency. The potency of the embryos depends on their size,
shape, orientation and the dislocation density at that site [28, 31]. Thus the em-
bryos become potent once they reach a critical size and attain a favorably oriented
shape that minimizes the combined interfacial and strain energies, i.e. ellipsoidal
shape [31], and also when there is a critical number of dislocations available at
the nucleation site [28]. It has also been reported that the stress-assisted trans-
formation initiates at the same nucleation sites where the athermal transformation
initiates [32].
After attaining potency, i.e. after a successful event of nucleation, the marten-
site embryos spontaneously grow with the aid of dislocation glide, in the case of
lath martensite [33, 34]. The growth of the martensite units gives rise to higher
level of internal stresses, which trigger autocatalysis, i.e. self nucleation of various
martensite domains in order to relax the internal stresses, as observed in the case
of lath martensite [35], plate martensite [3638] as well as in the case of isothermal
martensite [39]. Thus the transformation proceeds in three dierent stages, i.e.
nucleation, growth and autocatalysis.
2.5 Thermodynamic aspects
From a thermodynamic point of view, the reason for the embryos attaining potency
at the M
s
temperature, as explained above, can be attributed to the thermodynamic
driving force G
m
, i.e. the dierence between the Gibbs energies of austenite and
martensite. At T
0
temperature, where there is no chemical driving force available
as the Gibbs energies of the two phases are equal, the Gibbs energy barrier G
opposes the phase transformation. As the temperature is reduced to M

s
, during
quenching, the driving force becomes large enough such that it can overcome the
Gibbs energy barrier and hence the phase transformation takes place. Apart from
the chemical energy G
chem
, aroused due to quenching, the austenite-martensite
interface gives rise to an interfacial energy, i.e. the gradient energy G
grad
, and the
coherency of the two phases gives rise to internal stresses that in turn give rise to
elastic strain energy G
el
.
Based on the studies on nucleation and growth mentioned in Section 2.4, it can
be understood that there exist dierent thermodynamic driving forces that govern
the dierent stages of the martensitic transformation. Some works have showed that
a less driving force is needed for growth [40, 41] and autocatalysis [37] compared
to that required for nucleation. Thus martensite, once nucleated, could grow and
also give rise to autocatalysis above M
s
temperature. However Borgenstam et
al. [40] reported that the driving force for nucleation and growth are very close to
each other, although a less driving force is needed for growth compared to that for
nucleation.
In the case of stress-assisted martensite formation the externally applied me-
chanical energy G
appl
, depending on the nature of the load, either contributes to or
detracts to the thermodynamic driving force. Thus if the nature of load is favorable
for martensite formation, one can expect that M
s
is higher than M
s
, whereas M
s
is lower than M
s
if the nature of load is not favorable for martensite formation.
Moreover, for a given stress state, only those martensite units with the most favor-
able orientation with respect to the applied stress are formed. This eect is known
as the Magee eect [2426].
The thermodynamic aspects of both athermal and stress-assisted transforma-
tions, in terms of an internal variable that locally represents the extent of the
transformation, is schematically shown in Fig. 2.4. The internal variable can be
considered as a parameter that corresponds to the crystal structure of the two
phases, i.e. austenite ( = 0) and martensite ( = 1), as explained in the following
section.
2.6 Crystallography
The common plane, between the austenite and martensite, on which martensite
nucleates is called the habit plane [42]. Experimental results show that the habit
plane in the case of lath martensite is close to {111}
[43, 44] as well as {557}
[5].
As martensitic transformation occurs by a cooperative movement of atoms that
transforms the FCC structure in to the BCC or BCT structure, the habit plane, i.e.
the interface between the product and parent phases must be highly coherent. This
is also supported by the experimental observations that the habit plane is well-
dened crystallographically, i.e. there exist specic crystallographic orientation
2.6. CRYSTALLOGRAPHY 9
Figure 2.4: Schematic showing the thermodynamic aspects of athermal and stress-
assisted martensitic transformation. Gibbs energy curves in the absence of chemical
driving force, in the presence of chemical driving force only, as well as in the presence
of both chemical and mechanical driving forces.
relationships between austenite and martensite.
In order to understand the crystallography of martensite, phenomenological
theories were developed independently by Wechsler, Lieberman and Read and by
Bowles and Mackenzie [23]. Such phenomenological theories are based on the con-
cept that the habit plane should be an invariant plane, i.e. undistorted and unro-
tated plane [23]. Thus a deformation along the invariant plane that also satises
the coherency requirement is called an invariant plane strain, where twinning is an
example of such a deformation.
An austenite-martensite interface could be a coherent invariant habit plane if
the FCC and BCT are oriented in such a way that their dense-packed planes and
dense-packed directions are parallel to each other. Thus there exist specic crys-
tallographic orientation relationships between austenite and martensite, i.e. ori-
entation (rotation) of the habit plane of the martensite crystal with respect to a
particular crystallographic plane in the austenite crystal.
There exist four main austenite-martensite orientation relationships (OR),
namely Bain OR, Nishiyama-Wassermann (N-W) OR, Kurdjumov-Sachs (K-S) OR
and Greninger-Troiano (G-T) OR [23, 45, 46]. The K-S and N-W OR describe the
experimentally observed OR of martensite, whereas the G-T OR is intermediate
between the K-S and N-W OR [46]. The Bain OR is considered to be the basic OR
that can be used to explain the K-S and N-W OR, as will be explained below.
According to the Bain OR, an FCC austenite crystal ( = 0) can be transformed
to BCT martensite crystal ( = 1) by applying the Bain distortion (Bain strain), i.e.
a set of a compressive strain along one of the coordinate axes and two equal tensile
strains along the other two coordinate axes [23]. There exist three possibilities
to apply such a Bain strain on an FCC crystal and hence there exist three Bain
strains that give rise to three Bain variants (
1
,
2
,
3
), as shown in Fig. 2.5. The
OR between FCC and BCT crystals, according to Bain correspondence is shown in
Fig. 2.6.
Figure 2.5: Schematic showing the three dierent Bain variants [4].
Figure 2.6: Bain orientation relation showing the correspondence between FCC and
BCT crystals [23].
In an FCC crystal of austenite, there exist four close-packed planes, i.e. (111)
,
(111)
, (1 11)
, (11 1)
. Thus there exist four independent choices of the

{111}
plane. Both the K-S and N-W orientation relations are in agreement that
{111}
// {110}
. There exist three {110}
planes that could lie parallel to each of

the four {111}
planes. Therefore there exist 12 independent choices, i.e. 3 {110}
x 4 {111}
, of parallel planes of FCC and BCC along which martensite could form.
Figure 2.7: Illustration showing the 2 twin related K-S variants and 1 variant
accoriding to N-W OR. The three variants are examples of each of the three Bain
variants [47].
According to N-W OR, only one martensite variant could form along each of
these 12 parallel planes, as shown in Fig. 2.7, thereby giving rise to 12 N-W
martensite variants in total, i.e. 3 N-W martensite variants per each of the {111}
close-packed plane. According to K-S OR, two twin related martensite variants, as
shown in Fig. 2.7, could form along each of the 12 parallel planes, thereby giving
rise to 24 K-S variants in total, i.e. 6 K-S variants per each of the {111}
close-
packed plane, as shown in Fig. 2.8. More details regarding the above explanations
can be found in Refs. [45, 46].
It has been theoretically explained that the Bain strains (dilatation) coupled
with the shear and a rigid body rotation could give rise to the experimentally
observed habit planes [9, 45]. Thus the K-S and N-W OR, which are the exper-
imentally observed OR, are composed of Bain strains, shear and rotation. This
implies that each of the 24 K-S and 12 N-W martensite variants could be obtained
by applying Bain strain on the FCC lattice and thereafter by applying shear and
rotation. Thus it can be understood that all the above mentioned martensite vari-
ants could be grouped in to three distinct groups based on the applied Bain strains
(Bain distortion) and could be called Bain groups or Bain variants [4547].
Figure 2.8: A schematic of the K-S orientation relation [18].
Chapter 3
Phase-eld method
The modeling, performed more than a century ago by van der Waals, of a liquid-
gas system by means of density function that continuously varies over the inter-
face can be considered as the rst application of the phase-eld method. The
Ginzburg and Landaus work, approximately 50 years ago, to model superconduc-
tivity using a complex-valued order parameter can be considered to be the origin
of the concept of the order parameters, i.e. phase-eld variables. The micro-
scopic (continuous) theories proposed by Landau to describe the structural changes
occurring during a order-disorder transition by using mathematical expressions,
which are called Landau polynomials, are widely used in the modeling of ordering
transformations. The work by Cahn and Hilliard that proposed a thermodynamic
formulation that accounts for the gradients in thermodynamic properties in het-
erogeneous systems with diuse interfaces is usually regarded as the basis for the
present phase-eld models. Khachaturyan and his co-workers, in the 80s, have
developed Phase-eld Microelasticity theory, i.e. a phase-eld theory coupled with
elastic strain calculations based on the Eshelby approach, and by using Landau
polynomials to model martensitic transformations. Thereafter the order-disorder
transitions have been further explored by several researchers and their co-workers,
e.g. L.Q.Chen, Y.Wang, A.Artemev, A.Yamanaka, A.Finel, A.Saxena, T.Lookman,
S.Shenoy, K.Bhattacharya and V.Levitas. Another signicant contribution to the
phase-eld approach can be attributed to Langer, who developed a phenomeno-
logical single phase-eld model that is mainly used to model solidication. There-
after the solidication-type of models are well studied by several researchers, e.g.
J.Warren, B.Boettinger, R.Kobayashi, A.Karma, I.Loginova, J.Agren, G.Amberg
and N.Moelans. Phase-eld models are now widely used for simulating a wide range
of phenomena, e.g. microstructure evolution, solidication, precipitate growth and
coarsening, martensitic transformations, grain growth, solid state phase transfor-
mations, crack propagation and nucleation, dislocation motion [24]. More details
about the above mentioned history and the corresponding references can be found
in Ref. [4].
13
14 CHAPTER 3. PHASE-FIELD METHOD
Figure 3.1: A schematic of the diuse interface approach, i.e. continuous variation
in the internal properties across the interface [48].
Phase-eld method is a diuse interface approach, i.e. the internal properties
of a system undergoing a phase transformation vary continuously over the interface
between the two phases. The diuse interface approach can be explained with the
help of schematic Fig. 3.1 that shows the advancement of a domain wall at three
instances [48]. Each vertical rod represents the direction of magnetization of a plane
of atoms, whereas each row of rods represent the domain wall at an instance. On
either side of the center, indicated by an arrow in the middle, one can see that the
magnetic spin of atoms is dierent. It can also be seen that the spin direction varies
continuously over the interface. From time to time, it can be seen that the domain
wall (arrow) advances and also that the spin characteristics change continuously
over the interface. Hence the phase-eld method, which is based on the diuse
interface approach, is capable of predicting the continuous variation in the internal
properties over the interface between the two phases of a system that is undergoing
a phase transformation. Thus the advantage of the diuse interface approach is that
it avoids the need to explicitly track the position of the moving phase interface at
each instance. Thus phase-eld models are able to predict complex microstructural
evolutions.
During a phase transformation occurring in a material, the microstructure
evolves due to minimization of the Gibbs energy of the system. The Gibbs energy of
the system under consideration is described based on the underlying physics of the
phenomenon that has to be studied. Thus depending on the physical description of
the system, the Gibbs energy G could consist of dierent parts, e.g. bulk (G
bulk
),
interfacial (G
int
), elastic (G
el
), or any other applied external energy (G
appl
).
G = G
bulk
+ G
int
+ G
el
+ G
appl
(3.1)
The bulk part of the Gibbs energy G
bulk
corresponds to the bulk chemical proper-
ties of the two phases. The interfacial part of the Gibbs energy G
int
corresponds
to the properties of the interface. The elastic part of the Gibbs energy G
el
cor-
responds to the elastic strain energy that is generated inside the material due to
15
Figure 3.2: A schematic of the phase-eld approach.
the phase transformation, e.g. martensitic transformation, spinodal decomposition.
The applied external energy G
appl
corresponds to the energy generated due to an
externally applied force, e.g. mechanical force due to externally applied stresses,
magnetic forces due to an externally applied magnetic eld.
The Gibbs energy G is dependent on several internal variables (properties) that
describe the material characteristics, such as grain characteristics as well as the
composition of the material. In order to model the microstructure evolution, the
Gibbs energy G(, c) of the system is expressed as a mathematical function that
depends on the internal variables, and c. , a non-conserved quantity, governs the
characteristics of the grain (crystal) and c, a conserved quantity, is the composition
of the system. Such internal variables that describe the material characteristics are
eld variables that are continuous mathematical functions in time and space, i.e.
(x, y, z, t) and c(x, y, z, t), and are called phase-eld variables.
The values of the phase-eld variables vary between 0 and 1. In the case of a
single phase-eld approach used to study solidication, = 0 or 1 corresponds to
either of the two bulk phases, whereas a deviation from these two values corresponds
to the interface region between the two bulk phases. In the case of multi-phase-eld
approach used to study an order-disorder phase transformation, which gives rise to
an asymmetry in the resultant crystal structure such that each crystal orientation
of the product phase is distinct, one has to consider dierent phase-eld variables
that represent each of the above mentioned crystal orientations. Thus the number
of phase-eld variables will depend on the number of crystal orientations. Each
phase-eld variable will be equal to 0 in the product phase and equal to 1 in their
respective product phases.
The temporal, i.e. time-dependent, evolution of the microstructure is tracked
by solving a set of partial dierential equations (PDEs). Thus the solutions of
16 CHAPTER 3. PHASE-FIELD METHOD
the two PDEs yield the values of the internal variables at each time step and at
each point in the material and thereby predicts the changes in the microstructure.
A schematic of the phase-eld approach is presented in Fig. 3.2. The PDE that
tracks the changes in the non-conserved quantities, such as grain characteristics, is
usually referred as Cahn-Allen equation, whereas the PDE that tracks the changes
in the conserved quantities, such as composition, is usually referred as Cahn-Hilliard
equation [24]. Moreover both the equations contain several kinetic coecients that
govern the kinetics of the respective phase-eld variables. In the schematic Fig. 3.2,
the kinetic coecients are not presented in order to give a rather simple picture of
the phase-eld modeling concept. More detailed derivations and explanations can
be found in some of the reviews on phase-eld modeling [24].
Chapter 4
Phase-eld modeling of martensitic
transformations
The pioneering eorts in simulating the martensitic microstructure evolution could
be attributed to Khachaturyan [5] and his co-workers [6, 8, 9, 12]. The other signif-
icant contributions to this eld could be attributed to Bhattacharya [11], Saxena
and his co-workers [7, 49, 50], Levitas and his co-workers [10, 5153]. The eect of
the plastic deformation on the transformation, by coupling the phase-eld method
with the plasticity equation, has been studied by Tomita and his co-workers [13,14],
based on the work of Guo et al. [54].
The present work is based on the work by Khachaturyan et al. [5, 8, 9]. The
plastic deformation is also included, based on the work by Yamanaka et al. [13]. The
modeling of the transformation in polycrystals is based on the work by Yamanaka
et al. [14]. The phase-eld as well as the continuum mechanics equations are solved
by using the nite element method, which allows a straight forward and transparent
formulation of the equations.
The nite element method (FEM) is well established for the modeling of ther-
momechanical deformation. The basis of the method is to solve for the response
of a complicated body to imposed constraints by dividing it into smaller elements,
whose responses are more readily evaluated. Each element has a simple shape, such
as a triangle or quadrilateral for a 2D problem, or a tetrahedron for a 3D analysis.
The vertices of the elements are called nodes. The complete body is formed by
a mesh of elements, joined at nodes. Associated with each element is a stiness,
which is dependent upon the constitutive law of the material within the element.
By subjecting the element to a force, the overall response of the body is determined
by simultaneously solving for the node displacements such that the conditions of
equilibrium and boundary constraints are met. A detailed explanation about the
FEM can be found in Ref. [55].
All the single crystal computations are performed on a mesh with 50x50x50 grid
points and by using tetrahedral nite elements, whereas the polycrystal computa-
17
18
CHAPTER 4. PHASE-FIELD MODELING OF MARTENSITIC
TRANSFORMATIONS
tions are performed on a mesh with 150x150x150 grid points. The entire mathe-
matical framework is solved by using femLego software, a symbolic computational
tool [56]. femLego is a set of Maple procedures and fortran subroutines that can be
used to build complete fortran simulation codes for partial dierential equations,
with the entire problem denition done in Maple. OpenDX software is used for
plotting the images.
In the following sections, a brief overview of the phase-eld model developed
to study the morphology and transformation aspects, associated with athermal
martensitic transformation occurring in a single crystal as well as in a polycrys-
talline material and stress-assisted martensitic transformation occurring in a single
crystal, is presented. Einstein summation convention, i.e. summation implied over
repeated indices, is used to express all the equations except for the indices repre-
senting dierent martensitic domains.
A brief procedure explaining the multi-length scale modeling, i.e. by coupling
CALPHAD, ab-initio, phase-eld and experimental data, of martensitic transfor-
mation is also presented. A summary of the simulation procedures to study various
physical concepts, such as plastic deformation, nucleation and growth, thermody-
namic aspects and crystallography, associated with martensitic transformation is
also presented. More details can be found in the respective supplements.
4.1 Morphology and transformation
Athermal martensitic transformation, Supplement-1
In this part of the work, athermal martensitic transformation occurring in a single
crystal of Fe0.3%C steel is considered. A diusionless phase transformation, such
as martensitic transformation, can be modeled using the Cahn-Allen equation [57],
also known as the Time-Dependent Ginzburg-Landau (TDGL) kinetic equation [5].
The microstructure evolution can be simulated by predicting the time-dependent
variation of the phase-eld variable, which in turn is related to the minimization of
the Gibbs energy G of the system with respect to phase-eld variable
p
as:
p
t
=
q=v
q=1
L
pq
G
q
(4.1)
where
G
q
is a variational derivative that serves as a driving force for the forma-
tion of martensite denoted by the phase-eld variable
q
that is dependent on the
position vector r, v is the total number of martensite variants and L
pq
is a kinetic
parameter.
In the present work the Bain orientation relation is considered, as this is regarded
as the basic orientation relation that can be used to explain the K-S and N-W
orientation relations as explained in Section 2.6. According to Bain orientation
relation, explained in Section 2.6, the phase transformation of FCC austenite to
BCT martensite gives rise to three martensite variants (orientations). Thus in
4.1. MORPHOLOGY AND TRANSFORMATION 19
order to represent the three possible orientations, three phase-eld variables, i.e.
1
,
2
,
3
, need to be considered in the model and hence three phase-eld equations,
i.e. for p = 1, 2, 3 and v = 3 in Eq. (4.1), need to be solved at each time step. Thus
the three phase-eld variables correspond to the three Bain groups (Bain variants),
mentioned in Section 2.6.
From a thermodynamic point of view, as explained in Section 2.5, the Gibbs
energy consists of three parts:
G =
_
V
_
G
chem
v
+ G
grad
v
+ G
el
v
_
dV (4.2)
G
chem
v
corresponds to the chemical part of the Gibbs energy density of an un-
stressed system at the temperature under consideration. G
grad
v
is the extra Gibbs
energy density caused by the interfaces. The transformation of cubic austenite into
tetragonal martensite induces elastic strain energy density G
el
v
, into the material.
Chemical energy
The chemical part of the Gibbs energy density G
chem
v
, expressed as a Landau-type
polynomial [9, 58], is given by:
G
chem
v
(
1
,
2
,
3
) =
1
V
m
_
1
2
A
_
2
1
+
2
2
+
2
3
_
1
3
B
_
3
1
+
3
2
+
3
3
_
+
1
4
C
_
2
1
+
2
2
+
2
3
_
2
_ (4.3)
By considering the driving force G
m
, i.e. the dierence in the Gibbs energies
of austenite and martensite, and the Gibbs energy barrier G
terms in the above

equation, the coecients are modied [58] as: A = 32G
, B = 3A12G
m
and
C = 2A 12G
m
. G
is expressed as G
=
V
m
2
2
. V
m
is molar volume, is
thickness of the interface and relates to interfacial energy as explained below.
Gradient Energy
The gradient energy density term, G
grad
v
as presented in [9, 58] can be expressed as:
G
grad
v
=
p=v
p=1
ij
(p)
p
r
i
p
r
j
(4.4)
where r(x,y,z) is the position vector expressed in Cartesian coordinates.
ij
is the
isotropic gradient coecient matrix, i.e. diagonal tensor with all the elements equal
to =
9
2
V
m
16G
. is interfacial energy, V
m
is molar volume and G
is Gibbs energy
barrier.
20
TRANSFORMATIONS
Elastic energy
The internal elastic stresses developed in the material, as explained in Sections 2.2
and 2.5, give rise to elastic strain energy density G
el
v
. The elastic stress can be
calculated using the Microelasticity theory [5] as explained below. As mentioned in
Section 2.6, martensitic transformation gives rise to a change in the crystal structure
caused by Bain strains
00
ij
(r), which in turn induce stress-free transformation strains
0
ij
(r) in the material. The surrounding austenite matrix exerts strain
ij
(r) on the
martensite to resist the stress-free transformation strains and thereby induces elastic
strain
el
ij
(r), which in turn gives rise to stress
ij
(r), in the material. The actual
strain tensor
ij
(r) is linearly related to the local displacement vector u(r), which
is obtained by solving the mechanical equilibrium condition.
The elastic strain energy density term G
el
v
can be expressed as:
G
el
v
=
_

ij
(r)
0
ij
(r)
ij
(r)d
ij
(r) (4.5)
where
0
ij
(r) is the stress-free transformation strain,
ij
(r) is the actual strain and
ij
(r) is the stress, given by:
ij
(r) = c
ijkl
el
kl
(r) (4.6)
where c
ijkl
is the tensor of elastic modulii and
el
kl
(r) is the elastic strain, given by:
el
kl
(r) =
kl
(r)
0
kl
(r)
pl
kl
(r) (4.7)
where
pl
kl
(r) is the plastic strain.
The actual strain
ij
(r) is given by:
ij
(r) =
1
2
_
u
i
(r)
r
j
+
u
j
(r)
r
i
_
(4.8)
where u(r) is the local displacement vector obtained by solving the mechanical
equilibrium condition, expressed as:
ij
(r)
r
j
= 0 (4.9)
The displacement vector can then be used to calculate the actual strain
ij
(r),
which in turn can be used in calculating the local elastic strain energy density G
el
v
.
The stress-free transformation strain is given by:
0
ij
(r) =
p=v
p=1
p
(r)
00
ij
(p) (4.10)
where
00
ij
(p) is the Bain strain (Bain distortion) tensor, i.e. a set of a compressive
strain along one of the coordinate axes and two equal tensile strains along the other
two coordinate axes as explained in Section 2.6, given by:
00
ij
(1) =
_
_
3
0 0
0
1
0
0 0
1
_
_
,
00
ij
(2) =
_
_
1
0 0
0
3
0
0 0
1
_
_
,
00
ij
(3) =
_
_
1
0 0
0
1
0
0 0
3
_
_
(4.11)
where
3
=
c
t
a
c
a
c
is a compressive transformation strain,
1
=
a
t
a
c
a
c
is a tensile
transformation strain and are dened based on the lattice constants of austenite
(a
c
) and martensite (a
t
, c
t
). Thus, as shown in Fig. 2.5, martensite (Bain) variants-
1, 2 and 3 can be transformed from cubic austenite by applying Bain strains
00
ij
(1),
00
ij
(2) and
00
ij
(3) respectively.
The plastic strain comes into existence only when the stress exceeds the yield
limit, thereby initiating plastic deformation that acts as a relaxation of the stress, as
explained in Sections 2.2 and 2.3. von Mises criterion is employed to check whether
the stress has exceeded the yield stress
y
. According to von Mises criterion, plastic
deformation, i.e. yielding, occurs if Eq. (4.12) is satised.

2

2
y
=
1
2
_
(
xx

yy
)
2
+ (
yy

zz
)
2
+ (
zz

xx
)
2
_
+ 3
_
2
xy
+
2
yz
+
2
zx
_

2
y
0
(4.12)
where is von Mises equivalent stress.
In the absence of strain hardening eect, the plastic strain evolution equation
[13, 54, 58] can be expressed as:
pl
ij
(r)
t
= k
ijkl
G
shear
v
pl
kl
(r)
(4.13)
where
pl
ij
(r) is the local plastic strain, G
shear
v
is shear energy density and k
ijkl
is
plastic kinetic coecient expressed as:
k
ijkl
= kc
1
ijkl
(4.14)
where c
1
ijkl
denotes the compliance tensor and k is a parameter, which controls the
rate at which the stresses are relaxed by means of plastic deformation and is called
plastic relaxation rate.
The shear energy density G
shear
v
is given by:
G
shear
v
=
_
e
ij
(r)
e
0
ij
(r)
ij
(r)de
ij
(r)
=
_
e
ij
(r)
e
0
ij
(r)
c
ijkl
_
e
kl
(r) e
0
kl
(r)
_
de
ij
(r)
(4.15)
22
TRANSFORMATIONS
where e
ij
(r) and e
0
ij
(r) are the deviatoric actual strain tensor and deviatoric stress-
free transformation strain tensors given by:
e
ij
(r) =
ij
(r)
1
3
kk
(r)
ij
(4.16a)
e
0
ij
(r) =
0tot
ij
(r)
1
3
0tot
kk
(r)
ij
(4.16b)
where
0tot
ij
(r) is given by:
0tot
ij
(r) =
0
ij
(r) +
pl
ij
(r) (4.17)
The anisotropic elastic properties of dierent phases are taken in to account by
considering dierent tensors of elastic modulii c
ijkl
for dierent phases. In order
to consider dierent c
ijkl
an expression as shown in Eq. (4.18) is employed such
that for a given phase, i.e.
p
= 0 or
p
= 1, the corresponding tensors of elastic
modulii c
cubic
ijkl
or c
tetragonal
ijkl
are considered respectively whereas in the interface a
weighted c
ijkl
that depends on the weightage yielded by the phase-eld variable is
considered. The expressions of c
cubic
ijkl
and c
tetragonal
ijkl
are presented in Supplement-1.
c
ijkl
= c
cubic
ijkl
(1
p
) + c
tetragonal
ijkl

p
(4.18)
In all the gures, iso-surfaces (
p=1,2,3
= 0.5) of the three phase-eld variables,
i.e.
1
,
2
,
3
, that represent the three martensite (Bain) variants are plotted in
dierent colors, i.e. red, blue and green respectively. All the simulations in this
work are started with a pre-existing spherical martensite embryo of variant-1.
Stress-assisted martensitic transformation, Supplement-2
In this part of the work the above model in Supplement-1 is modied to study
the stress-assisted martensitic transformation occurring in a single crystal of Fe
0.3%C steel. In order to model the stress-assisted martensitic transformation, the
extra Gibbs energy density aroused due to the externally applied stress needs to be
included in the Gibbs energy of the system [8], as explained in Section 2.5. Hence
from a thermodynamic point of view, the Gibbs energy of a system undergoing
stress-assisted martensitic transformation consists of the following parts:
G =
_
V
_
G
chem
v
+ G
grad
v
+ G
el
v
+ G
appl
v
_
dV (4.19)
where G
chem
v
, G
grad
v
and G
el
v
are the chemical, gradient and elastic parts of the
Gibbs energy density, as explained in the above section. G
appl
v
is the extra Gibbs
energy density aroused due to the externally applied stress, expressed as:
G
appl
v
=
appl
ij

0
ij
(r) (4.20)
where
appl
ij
is the externally applied stress tensor, expressed by the Cauchy stress
tensor as:
appl
ij
=
_
_
xx

xy

xz
yx

yy

yz
zx

zy

zz
_
_
(4.21)
Eq. (4.2) is replaced with Eq. (4.19) and the entire model framework explained
in the above section corresponding to Supplement-1 is updated accordingly to model
the stress-assited martensite formation.
Multi-length scale modeling of martensitic transformations,
Supplement-3
In this part of the work, the above models in Supplements-1 and 2 are employed
to study the athermal as well as the stress-assisted martensitic transformations
occurring in a single crystal of stainless steel of type 301 with a composition of Fe
17%Cr7%Ni. Moreover an attempt is made to develop a multi-length scale model
by coupling the CALPHAD technique that is based on the fundamental principles
of thermodynamics, ab-initio at the atomistic level, phase-eld at the mesoscopic
level and the experimental works at the macro level.
The elastic constants of martensite (BCC) phase are obtained from ab-initio
calculations, the details of which are presented in Supplement-3. The rest of the
input parameters such as lattice constants, elastic constants of FCC phase, M
s
tem-
perature and interfacial energy are acquired from experimental data. The chemical
driving force available at M
s
is calculated using the CALPHAD technique.
Athermal martensitic transformation in polycrystalline material,
Supplement-4
In this part of the work, the model explained in Supplement-1 is modied to study
the athermal martensitic transformation occurring in a polycrystalline Fe0.3%C
steel. In order to simulate martensitic transformation in a polycrystalline material
that consists of several grains having random orientations, one needs to dene a
coordinate system in each grain. Each grain is represented by a local coordinate
system that is oriented at an angle with respect to the global coordinate system.
Thus each martensite variant needs to be rotated according to the grain orientation.
As each martensite variant is governed by the transformation strain tensor, the
rotation matrix that denes the grain orientation needs to be multiplied by the
stress-free transformation tensor.
To simulate the polycrystal in 3D, three Euler angles, , and are introduced
to describe the orientation of the grain in the global coordinate system based on
x-y-z convention. For the 3D case, Eq. (4.10) is modied as following:
24
TRANSFORMATIONS
0
ij
(r) =
p=v
p=1
Q
ik
(, , )Q
jl
(, , )
p
(r)
00
kl
(p) (4.22)
where the rotation matrix Q(, , ) is given by:
Q(, , ) =
_
_
coscos cossin + sinsincos sinsin + cossincos
cossin coscos + sinsinsin sincos + cossinsin
sin sincos coscos
_
_
(4.23)
Eq. (4.10) is replaced with Eq. (4.22) and the entire model framework explained
in the above section corresponding to Supplement-1 is updated accordingly to model
the athermal martensite formation in polycrystalline materials.
4.2 Plastic deformation aspects, Supplements-5 and 6
The plastic deformation aspects associated with the transformation can be studied
by studying the eect of plastic relaxation rate k on the transformation. Hence
dierent simulations are performed with dierent plastic relaxation rates k and
by considering the driving force for nucleation at M
s
. The model presented in
Supplement-1 is employed to perform the simulations by considering the athermal
martensitic transformation occurring in a single crystal of Fe0.3%C steel.
It can be observed from Eqs. (4.13) and (4.14) that k controls the plastic strain
rate and thereby quantitatively controls the extent of plastic deformation of the ma-
terial. Orowan equation states that the strain rate is proportional to the dislocation
density, Burgers vector and the dislocation velocity, which in turn is proportional
to the shear stress [59]. Thus by comparing Eq. (4.13) with the Orowan equation, it
can be understood that k is directly proportional to the dislocation mobility as well
as to the dislocation density. Hence by studying the eect of plastic relaxation rate,
the eect of dislocation density and dislocation mobility on the embryo potency,
morphology as well as on the overall transformation can be studied.
4.3 Nucleation and growth aspects, Supplement-5
Although nucleation is not simulated in the present work, several aspects associated
with the martensite nucleation are investigated by studying the eect of the marten-
site embryo potency on the transformation. The model presented in Supplement-1
is employed to perform the simulations by considering the athermal martensitic
transformation occurring in a single crystal of Fe0.3%C steel. As mentioned in
Section 2.4, the potency of the embryos depends on their size, shape, orientation
and the dislocation density at that nucleation site.
In order to study the eect of the dislocation density on the embryo potency,
the eect of the model parameter, i.e. plastic relaxation rate k in Eq. (4.14), on the
transformation is studied. Thus dierent simulations are performed with dierent
values of k. The relation between k and the dislocation density is explained above
in Section 4.2. In order to study the favorable size of the martensitic embryo, dif-
ferent simulations are performed with dierent embryo sizes. In order to study the
favorable shape of the martensitic embryo, dierent simulations are performed by
initiating the transformation (simulation) with a pre-existing spherical martensite
embryo and with an ellipsoidal embryo. In order to study the eect of the embryo
orientation on the transformation, three dierent simulations are performed by con-
sidering a pre-existing ellipsoidal shaped embryo with dierent orientations, i.e. the
longest axis of the ellipsoid being oriented along the three dierent coordinate axes.
4.4 Thermodynamic aspects, Supplement-6
As explained in Section 2.5, there exist dierent thermodynamic driving forces
that govern dierent stages of the transformation, such as driving forces required
for nucleation, growth and autocatalysis. The driving force for nucleation (G
nucl.
)
at M
s
is obtained by using Thermo-Calc software and TCFE6 database [60], which
is based on the CALPHAD method. For a given driving force G
nucl.
, the value of
the plastic relaxation rate k is determined, as explained above in Section 4.3, such
that the transformation occurs at the experimental M
s
, i.e. the embryo attains
potency at M
s
. The model presented in Supplement-1 is employed to perform the
simulations by considering the athermal martensitic transformation occurring in a
single crystal of Fe0.3%C steel.
The critical driving force required for growth (G
growth
) of martensite is de-
termined by starting the simulation with G
nucl.
and thereafter by reducing the
driving force in a step-wise manner until the growth completely stops. In order
to study the critical driving force required for autocatalysis (G
autocat.
), a similar
procedure as mentioned above is employed by ensuring that autocatalysis is avoided
at each step.
4.5 Crystallography, Supplement-6
The habit plane of martensite is determined by plotting the plane along which
the innitesimally small unit of martensite forms during the simulations. The
rotations of martensite variants, during the simulations, are studied by observing
the deviation of the martensite units from the above mentioned habit plane. The
concept of the invariant habit plane, as explained in Section 2.6, is also studied by
plotting a 2D section image of the 3D microstructure image and the corresponding
displacement plot. The model presented in Supplement-1 is employed to perform
the simulations by considering the athermal martensitic transformation occurring
in a single crystal of Fe0.3%C steel.
Chapter 5
Summary of results
5.1 Morphology and transformation
Athermal martensitic transformation, Supplement-1
The simulation results indicate that the martensitic microstructure consists of sev-
eral morphological mirrors, i.e. twinned variants marked by 1 and 2 as shown in
Fig. 5.1, as also observed in the experimental works [15]. It is observed from the
simulations, as shown in Fig. 5.2, that nucleation of several martensite variants,
i.e. autocatalysis, occurs as a consequence of the materials pursuit to minimize
the internal stresses. It is also observed that in the absence of plastic deforma-
tion and dilatation, the martensitic microstructure formation is completed earlier
compared to that in the presence of plastic deformation and dilatation, as shown
in Fig. 5.3. It can be understood that plastic deformation facilitates relaxation
Figure 5.1: Martensitic microstructure consisting of twinned martensite variants
[58].
27
28 CHAPTER 5. SUMMARY OF RESULTS
(a) (b) (c)
Figure 5.2: 3D simulation results with elastoplastic material and having anisotropic
elastic properties [58]. Snapshots at (a) t*=0 (b) t*=40 (c) t*=230.
Figure 5.3: Top view of the simulation results of athermal martensitic trans-
formation in single crystal. (a) Without dilatation and plastic deformation, at
non-dimensionless time t*=170. (b) With dilatation and plastic deformation, at
t*=230 [58].
of the stresses and hence reduces the need for minimization of stresses by means
of new martensite variant formation, i.e. autocatalysis. Thus it can be concluded
that both autocatalysis and plastic deformation are dierent modes of relaxing the
internal stresses generated due to the martensitic transformation.
From Fig. 5.3, it can be seen that martensite variants form in pairs, i.e. variant
pairs (1, 2), (1, 3) and (2, 3), in dierent directions. This pattern is observed in
both pure elastic material, i.e. Fig. 5.3a, as well as in the elastoplastic material, i.e.
Fig. 5.3b, and hence it can be understood that the formation of pairs of domains is
Figure 5.4: Results from athermal martensitic transformation in single crystal with
non-clamped boundary conditions, at t*=140 [58].
due to the tendency of the system to minimize its elastic strain energy. It is likely
that in a given direction, a variant combination is preferred in order to maximize
the energy minimization.
The simulation results indicate that the boundary conditions on the grain also
play a major role in the martensitic microstructure evolution. A clamped grain
with constrained boundary conditions gives rise to a muti-variant martensitic mi-
crostructure, whereas a non-clamped grain with unconstrained boundary conditions
gives rise to formation of a single martensite variant over the entire grain, as shown
in Fig. 5.4. In the case of a non-clamped system, the unconstrained boundaries
of the simulation domain accommodate the elastic stresses and hence there is no
urgent need for the material to form the autocatalysed martensite variants.
Stress-assisted martensitic transformation, Supplement-2
The results indicate that the role of stress states in martensitic microstructure
evolution is very crucial. The microstructure evolution is signicantly aected due
to the anisotropic loading conditions, as shown in Fig. 5.5. The dierent anisotropic
loading conditions signicantly aect the microstructures, as shown in Fig. 5.6.
Based on the microstructures obtained under dierent loading conditions, it can be
summarized that the variants that minimize the externally applied strain energy
and maximize the net driving force are favored, i.e. only variants with a favorable
orientation under a given stress state are formed, viz. Magee eect [2426].
The comparison of the von Mises equivalent plastic strain plots with the von
Mises equivalent stress plots indicates that the transformation induced plasticity
(TRIP) is an accomodation process to minimize the stresses aroused due to the
phase transformation, viz. Greenwood-Johnson eect [2426]. Fig. 5.7 shows a
stage during the microstructure evolution where the martensite variants have not
yet hit the (111)
plane, whereas Fig. 5.8 shows a subsequent evolution stage when

Figure 5.5: Top view of microstructure evolution under anisotropic loading
condition-8c, mentioned in Section 4.2 of Supplement-2, at dierent time steps
(a) t*=10 (b) t*=30 (c) t*=120.
Figure 5.6: Results from the simulations of stress-assisted martensitic transforma-
tion in single crystal, under anisotropic loading conditions (8a - 8f) mentioned in
Section 4.2 of Supplement-2. Snapshots at t*=120.
Figure 5.7: Microstructure, equivalent stress and equivalent plastic strain plots
obtained under uni-axial tensile loading at t*=85. (a) Side view of microstructure.
(b) Top view of the microstructure plotted on (111) plane. (c) Equivalent stress
corresponding to the microstructure plotted on (111) plane. (d) Equivalent plastic
strain corresponding to the microstructure plotted on (111) plane.
the martensite variants have hit the (111)
plane. It can be seen from Figs. 5.7

and 5.8 that as the martensite variants hit the (111)
plane from both sides, the

equivalent stress inside the plane increases, as can be observed in the corresponding
equivalent stress plot. Once the equivalent stress exceeds the yield limit, it intiates
plastic deformation as shown in the corresponding equivalent plastic strain plot,
thus indicating the initiation of the TRIP eect.
The simulation results, as shown in Fig. 5.9, indicate that the compressive
stresses applied by the martensite on austenite, due to the volume expansion caused
by the transformation, leads to stabilization of austenite and thereby gives rise to
higher content of retained austenite in between the martensite variants. From the
stress plot it can be seen that the boundaries of the martensite variants are in
Figure 5.8: Microstructure, equivalent stress and equivalent plastic strain plots
obtained under uni- axial tensile loading at t*=88. (a) Side view of microstructure.
(b) Top view of the microstructure plotted on (111) plane. (c) Equivalent stress
corresponding to the microstructure plotted on (111) plane. (d) Equivalent plastic
strain corresponding to the microstructure plotted on (111) plane.
Figure 5.9: Top view of microstructure and the corresponding equivalent stress
plotted on (111)
plane, obtained under a tri-axial anisotropic loading condition.

(a) Microstructure at t*=110. (b)von Mises equivalent stress of corresponding
microstructure at t*=110.
dark blue color, indicating that they exert a compressive stress on the surrounding
austenite, which is in accordance with Refs. [23, 61].
It has been observed in the simulations that the interaction between dierent
martensitic variants could also lead to higher internal stresses in the martensitic
interfaces and thereby gives rise to plastic accommodation, which is in good agree-
ment with the results of Marketz and Fischer [62]. By comparing the microstructure
with the stress plot in Fig. 5.10, it can be observed that the interfaces between
dierent martensitic variants are highly stressed regions and thereby lead to plastic
deformation in the interface regions, as can be seen in the plastic strain plot.
Multi-length scale modeling of martensitic transformations,
Supplement-3
The results indicate that a physically based multi-length scale model to simulate the
martensitic transformation in practically important materials such as stainless steels
is possible. The ab initio calculations of the elastic properties indicate that stainless
steels are elastically anisotropic, as shown in Fig. 5.11. The simulations predict the
athermal martensite formed in stainless steels to be of lath-type morphology, which
is in good agreement with experimental results [63]. It is observed that, most often,
all three martensite variants, i.e. all three Bain variants (groups), form in pairs as
Figure 5.10: Microstructure, equivalent stress and equivalent plastic strain plots ob-
tained under bi-axial compressive loading, plotted on (001) plane. (a) Microstruc-
ture at t*=37. (b) Equivalent stress corresponding to the microstructure at t*=37.
(c) Equivalent plastic strain corresponding to the microstructure at t*=37.
Figure 5.11: Calculated characteristic surfaces showing the Youngs modulus as a
function of crystallographic direction in a random FM Fe17%Cr7%Ni alloy. The
values on the color scale and on the axes are in GPa.
shown in Fig. 5.12. This result is dierent compared to the result obtained in the
case of carbon steel, where only two out of the three Bain variants are formed in
any given direction, as shown in Fig. 5.3b.
The simulation results of the stress-assisted martensite formation in stainless
steels under dierent stress states are similar to those obtained in the case of carbon
steels, as presented in Supplement-2.
Athermal martensitic transformation in polycrystalline material,
Supplement-4
The simulation results show that impingement of martensite variants on the grain
boundaries gives rise to higher stresses along the grain boundary. Such high stresses
are relaxed by martensite nucleation in the adjacent grain, viz. autocatalysis. The
results are in good agreement with the experimental observations that grain bound-
aries are probable sites for martensite nucleation [64], as shown in Fig. 5.13. It is
also observed that the orientation of a martensitic variant is governed by the orien-
tation of the grain, which is also in agreement with experimental results [64]. It can
be seen from Fig. 5.13 that the martensite variants, that grow in one grain, hit the
Figure 5.12: Results from the multi-length scale simulations of athermal martensitic
transformation in a single crystal of 301-type stainless steel, at t*=75.
grain boundaries and give rise to growth of martensite variants oriented dierently
in the adjacent grains. This phenomenon can be observed in the grains marked as
a and b in the simulated microstructure as well as in the grains marked as 1
and 2 in the experiementally obtained martensitic microstructure.
As observed in the case of athermal martensitic transformation in single crystal
explained above, plastic accommodation in polycrystalline material also reduces
the strain energy. Moreover, a coarser microstructure is obtained when plastic
deformation is considered in a polycrystalline material. Fig. 5.14 shows the 3D
simulation results of a clamped elastoplastic polycrystalline material.
In the case of a polycrystalline material, both clamped and non-clamped systems
resulted in a multi-variant martensitic microstructure. This is in contrast to that of
the results obtained in the case of a single crystal explained above, where clamped
grain resulted in a multi-variant microstructure and a non-clamped grain resulted
in the formation of a single martensite variant over the entire grain.
5.2 Plastic deformation aspects, Supplements-5 and 6
Based on the above results, it can be understood that plastic deformation plays
a signicant role in the martensitic transformation. Moreover, the simulations
performed to study the eect of plastic relaxation rate k on the transformation,
as mentioned in Section 4.2, indicate that the plastic relaxation rate aects the
morphology as well as the overall transformation.
5.2. PLASTIC DEFORMATION ASPECTS 37
Figure 5.13: Comparison of 2D simulation result from (a) an elastoplastic phase-
eld simulation of athermal martensitic transformation with that of (b) a grain
orientation map obtained from SEM-EBSD technique on a 301 steel sample that
has been subjected to athermal martensitic transformation [64].
Figure 5.14: Results from the simulations of athermal martensitic transformation
in an elastoplastic clamped polycrystalline material, at (a) t*=0 (b) t*=25 (c)
t*=300.
(a) (b)
Figure 5.15: Results from simulations with (a) low plastic relaxation rate (b) high
plastic relaxation rate [65].
A lower plastic relaxation rate results in a microstructure containing of many
martensite variants inside a single grain as shown in Fig. 5.15a. A higher plastic
relaxation rate results in a microstructure consisiting of a single martensite domain
growing across the entire grain as shown in Fig. 5.15b [65]. The above results are
in good agreement with the ndings of Levitas et al. [35].
The plastic relaxation rate k also aects the growth rate of martensite, as shown
in Fig. 5.16. Each sub-gure corresponds to snap shots, at t
= 50, from dierent

simulations with dierent k. It can be seen that under given conditions, Viz.
driving force and dimensionless time (t
= 50), the growth rate of a martensite

domain increases with increasing k. As k is directly proportional to the dislocation
mobility, as explained in Section 4.2, it can be understood that a higher dislocation
mobility gives rise to a higher growth rate of the domain [65]. This is in good
agreement with earlier observations that lath martensite grows with the aid of
dislocation glide [33, 34].
The von Mises equivalent plastic strain plot, as shown in Fig. 5.17, indicates
that the major plastic deformation occurs in the martensite compared to that of
austenite [65]. As a higher plastic strain implies a higher dislocation density, it
can be understood that martensite is a dislocation-rich structure, which is in good
agreement with the experimental observations of Morito et al. [27].
The plastic relaxation rate also aects the transformation by aecting the em-
bryo potency, as explained in the following section.
5.3 Nucleation and growth aspects, Supplement-5
The eect of plastic relaxation rate on the embryo potency is shown in Fig. 5.18.
The results indicate that for a given driving force there exists a critical plastic
5.3. NUCLEATION AND GROWTH ASPECTS 39
Figure 5.16: Results, at t
= 50 and mapped on (0 1 0) plane, from dierent

simulations with dierent k values, as mentioned below each sub-gure [65].
relaxation rate k
crit
, below which the embryo is impotent, i.e. reverse transforma-
tion from martensite to austenite. It can also be seen from Fig. 5.18 that as the
plastic relaxation rate is higher than the critical value, the embryo becomes potent
and grows. As the plastic relaxation rate is directly proportional to the disloca-
tion density, it can be construed that in order for the embryo to become potent, a
critical dislocation density is needed at the nucleation site [66]. The dependence of
embryo potency on the dislocation density is in good agreement with earlier report
by Cohen [28].
The simulations performed to study the eect of embryo potency on the trans-
formation, as mentioned in Section 4.3, indicate that the size, shape and orientation
of the martensite embryo aect the transformation. It becomes very dicult for a
small embryo to sustain the very high elastic stresses generated inside the material
and hence it shrinks and disappears completely. If the embryo is considerably big
such that it can sustain the elastic stresses, it becomes potent and starts to grow.
However the comparison of the critical size of the nucleus obtained from the sim-
(a) (b)
Figure 5.17: Illustration showing the maximum plastic strain concentrated in
martensite [65]. (a) Iso-surface of phase-eld variable (= 0.5) that corresponds
to a martensite domain, at t
= 35, mapped on (0 -1 1) plane with high plastic re-

laxation rate, i.e. k = 16.6k
crit
(b) The corresponding von Mises equivalent plastic
strain plot of the iso-surface shown in (a).
(a) (b) (c)
Figure 5.18: Iso-surfaces of phase-eld variable (= 0.5) obtained from 3D simula-
tions to study the eect of plastic relaxation rate (k) on the embryo potency [66].
Simulations performed with plastic relaxation rate (a) less than the critical value
(k < k
crit
) (b) equal to the critical value (k = k
crit
) (c) greater than the critical
value (k > k
crit
)
(a) (b) (c)
Figure 5.19: Iso-surfaces of phase-eld variable (= 0.5) obtained at t* = 80 from
three dierent simulations started with a pre-existing ellipsoidal embryo whose
major axis is aligned along (a) X-axis (b) Y-axis (c) Z-axis [66]. t* denotes dimen-
sionless time.
ulations with that of the size determined from earlier theoretical works [31] still
needs an in-depth analysis.
The simulations predict the favorable shape of the embryo as an ellipsoidal
shape, as can be seen in Fig. 5.18c, which minimizes the combined interfacial and
elastic strain energy of the embryo [31]. The results also indicate that a favorably
oriented embryo gives rise to a higher martensite volume fraction. The results of
three dierent simulations started with an ellipsoidal martensite embryo oriented
in dierent directions is shown in Fig. 5.19.
5.4 Thermodynamic aspects, Supplement-6
As mentioned in Section 4.4, the procedures to determine the dierent driving forces
required for growth and autocatalysis are shown in Figs. 5.20 and 5.21 respectively.
The results indicate that there exist three dierent critical (minimum) driving
forces, Viz. driving force required for nucleation (G
nucl.
), driving force required
for growth (G
growth
) and driving force required for autocatalysis (G
autocat.
),
which govern the martensitic transformation as well as the microstructure evolu-
tion. The simulations predict that G
growth
< G
autocat.
< G
nucl.
[65]. Thus
the results indicate that martensite, once nucleated, can grow above the M
s
tem-
perature. This is in agreement with the ndings of Borgenstam et al. [40]. However
the dierence in the temperatures corresponding to the nucleation and growth de-
termined from the present simulations is about 85 K, which is large compared to
that of 8 K reported by Borgenstam et al. It can be concluded that, regardless of
the absolute value of the temperature dierence, martensite growth could continue
above M
s
.
Figure 5.20: Illustration of the approach to predict the critical driving force for
growth [65].
Figure 5.21: Illustration of the approach to predict the critical driving force for
autocatalysis [65].
Figure 5.22: Prediction of initial habit plane [65].
5.5 Crystallography, Supplement-6
The model predicts the initial habit plane of the innitesimally small martensite
unit to be (111)
as shown in Fig. 5.22. In order to minimize the Gibbs energy of

the system, this martensite domain rotates away from the initial habit plane as the
transformation progresses, as shown in Fig. 5.23. On the formation of self nucleated
martensite units, the rate of rotation decreases as the energy minimization is now
achieved by means of autocatalysis. Thus the model predicts the nal habit plane of
the initially formed martensite domain, mentioned above, to be (211)
[65]. The
results are in agreement with the theoretical explanation that martensite variants
tend to rotate in order to minimize the strain energy [5, 9].
The rotations of martensite variants are studied in order to verify the theo-
retical explanation that dilatation coupled with shear and rotation could give rise
to the experimentally predicted habit planes [9], as explained in Section 2.6. The
dilatation is governed by the Bain strains, that are supplied as input parameters
for the simulations. The shear deformation is governed by the plasticity equation
(4.13). The rotation is governed by the minimization of the strain energy that
yields displacement, which in turn gives rise to the rotation of the variants [5, 9].
Fig. 5.24 shows a 2D section of the 3D athermal martensitic microstructure as
well as the corresponding displacement plot. In the displacement plot, red color
indicates maximum displacement and dark blue color indicates zero displacement.
By comparing the two images, it can be seen that the dark blue regions in the
displacement plot are the centers of the martensitic domains shown in the corre-
sponding microstructure image. This shows that the centers of the martensitic
domains are the regions where displacement is zero. Thus the habit plane, where
innitesimally small unit of martensite forms, is undistorted and unrotated as the
Figure 5.23: Simulation results showing the martensite domain rotating away from
the initial habit plane of (-1 1 1) [65]. (a) Initial habit plane at t
= 7 (b) Rotation
at t
= 45 (c) Rotation at t
= 150.
Figure 5.24: Illustration to explain the invariant habit plane concept. (a) 2D section
of the 3D athermal martensitic microstructure (b) Displacement plot of the corre-
sponding microstructure shown in (a). Red color indicates maximum displacement
and dark blue color indicates zero displacement.
displacement is zero. This is in agreement with the phenomenological theories of
martensite [23].
Both the phase-eld method and the Microelasticity theory [5] are based on
the continuum principles, i.e. principles that are used to describe a physical phe-
nomenon by using mathematical functions that are continuous functions in space
and time. Thus the phase-eld variables and the displacement eld, obtained by
the minimization of the elastic strain energy by using the Microelasticity theory
as explained in Eqs. (4.5) - (4.9) in Section 4.1, are continuous functions of the
coordinates in the lattice space. This ensures the crystal lattice continuity, i.e.
compatibility (coherency) of the austenite and martensite, by means of martensite
variant rotations as well as by means of the invariant habit plane [5, 9]. In simple
words, the martensite variants rotate and orient in such a way that the coherency
with austenite is maintained and also such that the strain energy is minimized.
Chapter 6
Concluding remarks and future
prospects
The results obtained in this work indicate that it is possible to model and to study
several aspects of the martensitic transformation using the phase-eld method in a
satisfactory way. Although the nucleation is not simulated in this work, the sim-
ulations performed to study the eect of embryo potency on the transformation
indicate that the size, shape and orientation of the embryo as well as the dislo-
cation density at the nucleation site govern the martensitic transformation. The
simulations predict that martensite, once nucleated, could grow as well as lead to
autocatalysis, above M
s
. Although this result is in agreement with experimental
results [40], the experiments indicate that martensite could only grow up to a tem-
perature that is slightly above M
s
, whereas the simulations predict that martensite
could grow well above M
s
and hence needs to be studied in detail by performing
more experiments as well as simulations.
The experimental observations that the martensitic microstructures consist of
several twinned martensite variants are observed in the simulated microstructures
as well. Autocatalysis, i.e. nucleation of several martensite variants, plastic defor-
mation as well as the non-clamped boundary conditions act as relaxation measures
of the stresses developed, due to the martensitic transformation, inside the mate-
rial. The microstructures obtained in a polycrystalline material indicate that the
martensite variants growing in one grain, when impinge up on the adjacent grains,
could initiate martensite nucleation in the adjacent grains, where the grain bound-
aries act as nucleation sites. The study of plastic deformation aspects show that
the martensitic microstructure is signicantly aected by the plastic relaxation rate,
which corresponds to the dislocation mobility and density.
The model predicts the initial habit plane of martensite in agreement with the
experimental results and predicts that the variants rotate in order to minimize the
strain energy. If the rotations of various variants could also be plotted along with
the Bain variants shown in the present gures, it might be possible to predict some
47
48 CHAPTER 6. CONCLUDING REMARKS AND FUTURE PROSPECTS
Figure 6.1: Variation of M
s
with externally applied uni-axial tensile stress.
more aspects of martensite crystallography and hence this aspect needs an in-depth
study.
The simulations of the stress-assisted martensitic microstructure evolution un-
der dierent loading conditions are appealing and indicate that the microstructure
evolution is signicantly aected by the stress states and also that the anisotropic
loading conditions give rise to a higher level of anisotropy in the microstructure.
The results clearly depict the Magee eect. In future, the simulations can be supple-
mented by experimental works. Moreover, some practically important parameters
such as M
s
temperature and the volume fraction of the martensite formed can also
be predicted from the simulations, as shown in Figs. 6.1 and 6.2 respectively.
The study of the von Mises plastic strain plots of the microstructures enables
to understand the TRIP eect as well as the interaction between martensite vari-
ants. The TRIP eect can also be understood by studying the variation of von
Mises equivalent plastic strain with respect to the martensite volume fraction, as
shown in Fig. 6.3. It can be seen that the plastic strain drastically increases once
the martensitic transformation starts and thereafter increases with the increasing
martensite volume fraction, indicating the initiation of the transformation induced
plasticity (TRIP) eect.
As a continuation of the present work, the model can be improved further by
including the strain hardening eect in the plasticity equations. The plastic de-
49
Figure 6.2: Variation of martensite volume fraction as the microstructure evolves
under applied uni-axial tensile stress.
formation model can also be further improved by considering the crystal plasticity
framework, which might make it feasible to study the martensite formation along
dierent slip planes.
Based on the multi-length scale modeling results, it can also be concluded that it
is possible to build a physically based multi-length scale model, by coupling dier-
ent computational methods as well as experimental works, to study the martensitic
transformations. However in order to further improve the multi-length scale model,
the macroscopic properties of the material also need to be studied by investigat-
ing the response of the dierent simulated martensitic microstructures to various
external loading conditions. Such attempts might also contribute to the Materials
Genome initiative, which is now one of the most widely discussed applied research
topics among the scientic community.
The experimental studies on martensite morphology and crystallography using
advanced characterization techniques, such as EBSD, FIB etc., seem very interest-
ing. Moreover the recent studies of 3D reconstruction of martensitic microstruc-
tures [67] are exciting. Hence it could be interesting to compare such 3D exper-
imental results with the 3D phase-eld modeling results obtained in the present
work, in order to bridge the gaps between the two realms, viz. experiments and
models.
50 CHAPTER 6. CONCLUDING REMARKS AND FUTURE PROSPECTS
Figure 6.3: Variation of mean equivalent plastic strain with martensite volume
fraction obtained under applied uni-axial tensile stress.
Thus with a tone of cautious optimism, it can be concluded by saying that
phase-eld method can be a useful qualitative tool in understanding several aspects
associated with the martensitic transformations, provided the underlying physics is
taken care of in an appropriate manner.
Acknowledgements
First of all I would like to express my sincere gratitude to my principal supervisor
Prof. John gren for giving me this opportunity to work with one of the intriguing
and exciting research topics in the area of Physical Metallurgy, for his supervision
and for all the interesting discussions. I would also like to thank my another
supervisor Assoc. Prof. Annika Borgenstam for the critical assessment of my work
that has greatly added value to the work. I want to express my thanks to all the
other collaborators, Prof. Gustav Amberg, Mr. Amer Malik, Dr. Peter Hedstrm,
Dr. Vsevolod Razumovskiy, Dr. Pavel Korzhavyi and Prof. Andrei Ruban. My
special thanks to Prof. Emer. Mats Hillert for the interesting discussions and ideas.
Thanks to Dr. Lars Hglund and Dr. Minh Do Quang for all the help with the
computer related aspects. I would also like to thank Prof. Hasse Fredriksson, whom
I always consider as my rst teacher in the eld of Materials research. Thanks to
Prof. Staan Hertzman for the motivation to work on stainless steels. My thanks
to all the other colleagues at MSE that have made all these years pleasant and
enjoyable.
I would like to thank the VINN Excellence Center Hero-m, Vinnova, Swedish
Industry and KTH Royal Institute of Technology for providing nance for perform-
ing this work. Thanks to SNIC for providing the computer resources as well as
to the STT foundation, Wallenberg foundation and Jernkontoret for providing the
travel grants to attend various conferences.
I would also like to thank Prof. A.G. Khachaturyan, Dr. Avadh Saxena, Dr.
Turab Lookman, Dr. Alfonse Finel, Prof. Greg Olson, Prof. S.Q. Shi and Prof.
Kaushik Bhattacharya for all the discussions on martensite and phase-eld mod-
eling, during dierent conferences, which have greatly helped me in understanding
this subject.
Finally, I would like to express my deepest gratitude to the two most wonderful
persons in my life, my mother and aunt, for all their support and encouragement
in all the phases of my life. Thanks to my best friend Sudheer for his support and
friendship. Thanks to my friends, Sathees, Kavitha and Mitra for their support
and making my life in Sweden quite pleasant. A special thanks to Reshu for the
motivation to complete this work on the martensitic transformation in a martensitic
manner.
51
Bibliography
[1] P.C. Clapp, Journal De Physique IV 5 (1995) C8-11.
[2] W.J. Boettinger, J.A. Warren, C. Beckermann, A. Karma, Annual Review of
Materials Research 23 (2002) 163.
[3] L.Q. Chen, Annual Review of Materials Research 32 (2002) 113.
[4] N. Moelans, B. Blanpain, P. Wollants, CALPHAD 32 (2008) 268.
[5] A.G. Khachaturyan, Theory of Structural Transformations in Solids, Dover
Publications Inc., New York (2008).
[6] Y. Wang, A.G. Khachaturyan, Acta Materialia 45 (1997) 759.
[7] A. Saxena, A.R. Bishop, S.R. Shenoy, T. Lookman, Computational Materials
Science 10 (1998) 16.
[8] A. Artemev, Y. Wang, A.G. Khachaturyan, Acta Materialia 48 (2000) 2503.
[9] A. Artemev, Y. Jin, A.G. Khachaturyan, Acta Materialia 49 (2001) 1165.
[10] V.I. Levitas, D.L. Preston, Physical Review B 66 (2002) 134206.
[11] K. Bhattacharya, Microstructure of Martensite - How it forms and how it
gives rise to the shape-memory eect, Oxford University Press Inc., New York
(2003).
[12] Y. Wang, A.G. Khachaturyan, Materials Science and Engineering A 438-440
(2006) 55.
[13] A. Yamanaka, T. Takaki, Y. Tomita, Materials Science and Engineering A 491
(2008) 378.
[14] A. Yamanaka, T. Takaki, Y. Tomita, International Journal of Mechanical Sci-
ences 52 (2010) 245.
[15] G. Krauss, A.R. Marder, Metallurgical Transactions 2 (1971) 2343.
53
54 BIBLIOGRAPHY
[16] S. Morito, X. Huang, T. Furuhara, T. Maki, N. Hansen, Acta Materialia 54
(2006) 5323.
[17] G. Krauss, Materials Science and Engineering A 273-275 (1999) 40.
[18] S. Morito, H. Tanaka, R. Konishi, T. Furuhara, T. Maki, Acta Materialia 51
(2003) 1789.
[19] S. Morito, H. Yoshida, T. Maki, X. Huang, Materials Science and Engineering
A 438-440 (2006) 237.
[20] A. Shibata, S. Morito, T. Furuhara, T. Maki, Scripta Materialia 53 (2005) 597.
[21] A. Shibata, S. Morito, T. Furuhara, T. Maki, Acta Materialia 57 (2009) 483.
[22] P. Kolmskog, Work performed as part of his doctoral thesis, To be published.
[23] Z. Nishiyama, Martensitic Transformation, Academic Press, New York (1978).
[24] M. Cherkaoui, M. Berveiller, X. Lemoine, International Journal of Plasticity
16 (2000) 1215.
[25] F.D. Fischer, G. Reisner, E. Werner, K. Tanaka, G. Cailletaud, T. Antretter,
International Journal of Plasticity 16 (2000) 723.
[26] R.F. Kubler, M. Berveiller, P. Buessler, International Journal of Plasticity 27
(2011) 299.
[27] S. Morito, J. Nishikawa, T. Maki, ISIJ International 43 (2003) 1475.
[28] M. Cohen, Proceedings of International Conference on Martensitic Transfor-
mations (1986) 16.
[29] M. Suezawa, H.E. Cook, Acta Metallurgica 28 (1980) 423.
[30] J.W. Brooks, M.H. Loretto, R.E. Smallman, Acta Metallurgica 27 (1979) 1839.
[31] M. Cohen, Transactions Metallurgical society AIME (1958) 171.
[32] G.B. Olson, M. Cohen, Journal of Less-Common Metals 28 (1972) 107.
[33] K.E. Easterling, A.R. Thlen, Acta Metallurgica 28 (1980) 1229.
[34] G. Thomas, B.V.N. Rao, Proceedings of Conference on Martensitic Transfor-
mations, Kiev (1978) 57.
[35] V.I. Levitas, A.V. Idesman, G.B. Olson, E. Stein, Philosophical Magazine A
82 (2002) 429.
[36] E.S. Machlin, M. Cohen, Transactions AIME 191 (1951) 746.
BIBLIOGRAPHY 55
[37] J.C. Bokros, E.R. Parker, Acta Metallurgica 11 (1963) 1291.
[38] J.R.C. Guimaraes, Materials Science and Engineering 95 (1987) 217.
[39] V. Raghavan, in: G.B. Olson, W.S. Owen (Eds.), Martensite, ASM Interna-
tional, Ohio (1992) 197.
[40] A. Borgenstam, M. Hillert, J. gren, Acta metallurgica materialia 43 (1995)
945.
[41] H. Warlimont, Transactions Metallurgical Society AIME 224 (1962) 495.
[42] J.A. Klostermann, Journal of Less-Common Metals 28 (1972) 75.
[43] P.M. Kelly, A. Jostsons, R.G. Blake, Acta Metallurgica Materialia 38 (1990)
1075.
[44] B.P.J. Sandvik, C.M. Wayman, Metallurgical Transactions A 14 (1983) 809.
[45] Z. Guo, C.S. Lee, J.W. Morris Jr., Acta Materialia 52 (2004) 5511.
[46] P.P. Suikkanen, C. Cayron, A.J. DeArdo, L.P. Karjalainen, Journal of Mate-
rials Science Technology 27 (2011) 920.
[47] J.W. Morris Jr., ISIJ International 51 (2011) 1569.
[48] J.W. Cahn, Acta Metallurgica 8 (1960) 554.
[49] S.R. Shenoy, T. Lookman, A. Saxena, A.R. Bishop, Physical Review B 60
(1999) R12537.
[50] R. Ahluwalia, T. Lookman, A. Saxena, R.C. Albers, Acta Materialia 52 (2003)
209.
[51] V.I. Levitas, D.W. Lee, Physical Review Letters 99 (2007) 245701.
[52] A.V. Idesman, J.Y. Cho, V.I. Levitas, Applied Physical Letters 93 (2008)
043102.
[53] V.I. Levitas, D.W. Lee, D.L. Preston, International Journal of Plasticity 26
(2010) 395.
[54] X.H. Guo, S.Q. Shi, X.Q. Ma, Applied Physics Letters 87 (2005) 221910.
[55] Y.C. Fung, P. Tong, Classical and computational solid mechanics, World Sci-
entic publishing Co., Singapore (2001).
[56] G. Amberg, R. Tnhardt, C. Winkler, Mathematics and Computers in Simu-
lation 49 (1999) 149.
[57] S.M. Allen, J.W. Cahn, Acta metallurgica 27 (1979) 1085.
56 BIBLIOGRAPHY
[58] H.K. Yeddu, A. Malik, J. gren, G. Amberg, A. Borgenstam, Acta Materialia
60 (2012) 1538.
[59] D. Hull, D.J. Bacon, Introduction to Dislocations, fth ed., Elsevier Ltd.,
Oxford (2011).
[60] J.O. Andersson, T. Helander, L. Hglund, P.F. Shi, B. Sundman, CALPHAD
26 (2002) 273.
[61] I. Karaman, H. Sehitoglu, H.J. Maier, M. Balzer, Metallurgical and Materials
Transactions A 29A (1998) 427.
[62] F. Marketz, F.D. Fischer, Modelling and Simulation in Materials Science and
Engineering 2 (1994) 1017.
[63] B.P.J. Sandvik, H.O. Martikainen, V.K. Lindroos, Scripta Metallurgica 18
(1984) 81.
[64] R. Naraghi, P. Hedstrm, A. Borgenstam, Steel Research International 82
(2011) 337.
[65] H.K. Yeddu, A. Borgenstam, P. Hedstrm, J. gren, Materials Science and
Engineering A 538 (2012) 173.
[66] H.K. Yeddu, A. Borgenstam, J. gren, Journal of Alloys and Compounds, In
Press (2012), doi:10.1016/j.jallcom.2012.01.087.
[67] S. Morito, Y. Edamatsu, K. Ichinotani, T. Ohba, T. Hayashi, Y. Adachi, T.
Furuhara, G. Miyamoto, N. Takayama, Journal of Alloys and Compounds, In
Press (2012), doi:10.1016/j.jallcom.2012.02.004.

Martensit

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Martensit

Uploaded by

Copyright:

Available Formats

Martensitic Transformations in Steels

opposes the phase transformation. As the temperature is reduced to M

[43, 44] as well as {557}

. Thus there exist four independent choices of the

. There exist three {110}

planes that could lie parallel to each of

planes. Therefore there exist 12 independent choices, i.e. 3 {110}

terms in the above

plane, whereas Fig. 5.8 shows a subsequent evolution stage when

plane. It can be seen from Figs. 5.7

plane from both sides, the

plane, obtained under a tri-axial anisotropic loading condition.

= 50, from dierent

= 50), the growth rate of a martensite

= 50 and mapped on (0 1 0) plane, from dierent

= 35, mapped on (0 -1 1) plane with high plastic re-

as shown in Fig. 5.22. In order to minimize the Gibbs energy of

You might also like